Grzegorz Nosek

@FedeDP no, it's my garbage PR to use the Windows CI ;) Once I clean it all up and I'm happy with the changes, I'll submit them as proper PRs...

Huge pages are an optimization, we only need to worry about the smallest page size per architecture :)

Because that's the mmap granularity (if you mmap a multiple of a huge page, you can get one big page instead of a bunch of smaller ones but having huge...

Technically yes, though I don't think we should worry too much about ia64 in the first place :D

Well, on hexagon the minimum buffer size is still two pages, not our fault the pages are huge 🤷

One, two, many, lots ;)

If you have a specific use for it, sure. But also, can this happen e.g. early in init() instead? We could make sure that nothing too complex happens before init()...

> "if we today are able to push to userspace 1mln fd-only events, we could instead push up to 2mln fd-only events" This hits even harder :) Optimism is awesome...

Marking as wip since we need to handle modern_bpf as well

Hmm but what is it that we're actually discussing here? * Switching from sys_enter/sys_exit to per-syscall tracepoints? * Adding some infrastructure so that the above can be done in a...